Land Acknowledgement

With our deepest respects to the Tongva and Serrano Peoples, past, present, and emerging.

Stanford Policing Data

https://openpolicing.stanford.edu/

Stanford Policing Data

Just under 100 datasets. https://openpolicing.stanford.edu/data/

Why this project?

• Engaging questions (for me and for students)
• Goldilocks level of data wrangling
• Each student can work with a different dataset
• Ability (need!) to work with SQL

Step 1: get data

con <- dbConnect(
  MySQL(), host = "XXX", user = "XXX",
  password = "XXX", dbname = "XXX")

raleigh_df <- DBI::dbGetQuery(con, "SELECT * FROM NCraleigh") 

Step 2: data viz

raleigh_df %>%
  
  # remove missing data
  filter(!is.na(sex) & !is.na(race)) %>%
  
  # use group_by and summarize to count number of stops per category
  group_by(sex, race) %>%
  summarize(count = n()) %>%
  ungroup() %>%
  
  # find the percentage of age/race stops
  mutate(percentage = round(prop.table(count), digits = 2)) %>%

Step 2: data viz

… continued

  # plot percentages
  ggplot(mapping = aes(x = sex, y = percentage, 
                       fill = race,
                       label = scales::percent(percentage))) +
  geom_bar(position = "dodge", stat = "identity") +
  
  # adjust labels
  geom_text(position = position_dodge(width = .9),    
            vjust = -0.5,   
            size = 3) + 
  scale_y_continuous(labels = scales::percent) +

  # provide labels
  ggtitle("Race & gender breakdown, % out of total")

Step 2: data viz

Step 2: advanced data viz

Accessing, joining, and wrangling all the datasets.

Step 2: advanced data viz

With facet_geo

Step 3: modeling search

n.b., all the observations were traffic stops, so we can’t model demographics of who was pulled over (model search instead).

raleigh_search <- glm(formula = search  ~ age * race, 
    family = "binomial", data = raleigh_df, 
    subset = (race %in% c("black", "white")))

raleigh_search %>% tidy()

## # A tibble: 4 x 5
##   term          estimate std.error statistic  p.value
##   <chr>            <dbl>     <dbl>     <dbl>    <dbl>
## 1 (Intercept)   -1.90     0.0240      -78.9  0.      
## 2 age           -0.0327   0.000759    -43.1  0.      
## 3 racewhite     -0.785    0.0402      -19.5  5.39e-85
## 4 age:racewhite  0.00200  0.00125       1.60 1.10e- 1

Step 3: data viz of model

Language around sensitive data

Data Feminism https://datafeminism.io/

Equity vs Equality

Image credit: Interaction Institute for Social Change | Artist: Angus Maguire.

Traffic stops

What are the systems that have created structural racial discrepancies in the US?

• Redlining: systematic exclusion of people of color from obtaining mortgages (Federal lending programs)
• Home ownership is lower for people of color
• Housing discrimination continues today and creates racial differences in where people live and work.

Traffic stops

Pierson at al. “A large-scale analysis of racial disparities in police stops across the United States” (Nature Human Behaviour, 2020)

Our results indicate that police stops and search decisions suffer from persistent racial bias and point to the value of policy interventions to mitigate these disparities.

While the research shows that stops are neither equal nor equitable, a discussion on equity belongs in the conversation around racial discrepancies in traffic stops.

8am weekly zoom calls

The undergraduates who did the complete analysis and wrote the report are Pomona College students: Amber Lee (‘22), Arm Wonghirundacha (‘22), Emma Godfrey (‘21), Ethan Ong (‘21), Ivy Yuan (‘21), Oliver Chang (‘22), and Will Gray (‘22).

Thank you!